Pete Zaitcev (zaitcev(a)redhat.com) said:
I seem to have figured this out. The udev is not implicated at all,
because the kernel will not deliver any hotplug events until the
pcmcia-socket-startup was run (so udev cannot do anything about it).
Harald suggested to look at scsi_replay() in /sbin/start_udev.
I did, and while interesting, it seems like an overkill.
I don't know if we want to enable pcmcia sockets that early.
We do not boot from CF-in-PCMCIA adapter, right? My main concern
is that anything running so early is hard to debug.
Bill, how about adding a script to pcmciautils (see appended)?
Looks like overkill for a configuration we don't build.
Realistically, there will probably be PCI replay stuff soon,
which should cover this.
Bill